Micro-grid energy storage full life cycle optimization operation method and system

Abstract

The present application relates to the technical field of micro-grid, and discloses a kind of micro-grid energy storage full life cycle optimization operation method and system.The method includes: with the minimum full life cycle operating cost as target to build micro-grid scheduling model, reconstructs Markov decision process according to operating day;Respectively build energy storage cumulative life loss state value function and state of charge value function, and carry out piecewise linear approximation;With the sum of current operating day cost and two kinds of value functions as optimization goal, solve scheduling decision day by day;According to scheduling decision, calculate slope sample value and iterative update until value function converges;Using the value function after convergence obtains final scheduling decision of each operating day.The present application converts long-period optimization into phased decision problem, quantifies the influence of current behavior on subsequent cost through value function, and uses multi-precision acceleration training strategy to improve convergence efficiency while ensuring accuracy, effectively reducing the complexity of model solving under long time scale.

Description

Technical Field

[0001] This invention relates to the field of microgrid technology, specifically to a method and system for optimizing the operation of microgrid energy storage throughout its entire lifecycle. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] In existing microgrids, energy storage can coordinate the output of new energy sources, load demand, and interaction with the external power grid through charging and discharging, making it an important resource for improving the economic efficiency and energy regulation capabilities of microgrid operations. However, energy storage experiences lifespan loss during operation, and load, electricity prices, and wind and solar output vary with the operating phase, resulting in differences in the operational benefits generated by using energy storage on different operating days.

[0004] Existing energy storage optimization scheduling methods typically limit the intensity of energy storage use by converting lifetime loss into operating costs or by setting constraints such as discharge depth, cycle count, and cumulative lifetime loss limits. These methods are mostly aimed at short-term or single-day scheduling. However, the charging and discharging decisions of the current operating day not only affect the operating cost of that day but also change the available lifetime resources and energy state on subsequent operating days. Therefore, existing energy storage optimization scheduling methods are difficult to allocate limited lifetime resources in a coordinated manner based on the operating value of different operating days. If a scheduling model covering the entire life cycle of energy storage is directly constructed, the energy storage lifetime loss and state of charge will continuously be transmitted and coupled between operating days. As the scheduling cycle lengthens, the model size and solution complexity increase significantly. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for optimizing the entire lifecycle operation of microgrid energy storage. This invention transforms a long-term optimization problem into a staged decision-making problem, utilizing value functions to quantify the impact of current energy storage behavior on subsequent operating costs, thereby reducing the solution difficulty caused by the increasing model size over long time scales.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: Firstly, a method for optimizing the entire lifecycle operation of microgrid energy storage is provided, including: Obtain microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total lifecycle operating cost; The microgrid scheduling model is reconstructed into a Markov decision process according to the operating day; The cumulative lifetime loss state value function and the charge state value function for each operating day are constructed respectively, and piecewise linear approximations are performed on the two value functions to determine the initial slope of each segment. According to the order of operating days, an optimization objective is constructed based on the operating cost, the cumulative lifetime loss state value function, and the charge state value function of the current operating day, and the scheduling decision for the current operating day is solved. The initial state of the next operating day is updated based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round. Based on the scheduling decisions for each operating day of this round, the slope sampling values ​​of the two value functions are calculated respectively, and the slope of each segment is updated accordingly; the scheduling decisions for all operating days and subsequent slope updates are repeated until the two value functions meet the convergence condition; a multi-precision accelerated training strategy is adopted in the iterative update process of the two value functions; in the initial update stage, multiple consecutive target segments are divided into several sub-intervals, so that multiple segments in the same sub-interval share the same slope, and the shared slope of the sub-intervals is updated according to the scheduling results of the operating days; after obtaining the overall trend of the value functions, the segments in each sub-interval are restored to independent segments, and the convergence slope of the corresponding sub-interval is used as the initial slope of each independent segment, and slope sampling and updating are performed respectively; Using the two converged value functions, the scheduling decision is executed again in the order of the operating days to obtain the final scheduling decision for each operating day.

[0007] Secondly, a microgrid energy storage full lifecycle optimized operation system is provided, including: The model building module is configured to: acquire microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total life cycle operating cost; The model reconstruction module is configured to reconstruct the microgrid scheduling model into a Markov decision process according to the operating day. The daily decision-making module is configured to: construct the energy storage cumulative lifetime loss state value function and the charge state value function for each operating day, and perform piecewise linear approximation on both functions to determine the initial slope of each segment; construct an optimization objective based on the operating cost, energy storage cumulative lifetime loss state value function, and charge state value function of the current operating day according to the operating day sequence, and solve the scheduling decision for the current operating day; update the initial state of the next operating day based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round; The iterative training module is configured to: calculate the slope sampling values ​​of the two value functions according to the scheduling decisions for each running day of the round, and update the slope of each segment accordingly; repeat the scheduling decisions and subsequent slope updates for all running days until the two value functions meet the convergence conditions; adopt a multi-precision accelerated training strategy during the iterative update of the two value functions; in the initial update phase, divide multiple consecutive target segments into several sub-intervals, so that multiple segments within the same sub-interval share the same slope, and update the shared slope of the sub-intervals according to the scheduling results of the running days; after obtaining the overall trend of the value functions, restore the segments within each sub-interval to independent segments, and use the convergence slope of the corresponding sub-interval as the initial slope of each independent segment, and perform slope sampling and updating respectively; The final scheduling module is configured to: use the two converged value functions to execute the scheduling decision again in the order of the running days to obtain the final scheduling decision for each running day.

[0008] Thirdly, an electronic device is also provided, comprising: Memory is used to store computer-readable instructions in a non-transitory manner. Processor, for executing the computer-readable instructions, When the computer-readable instructions are executed by the processor, they perform the method described in the first aspect above.

[0009] Fourthly, a computer-readable storage medium is provided having a program stored thereon that, when executed by a processor, implements the method described in the first aspect above.

[0010] Fifthly, a computer program product is provided, employing the following technical solution: A computer program product includes software code, wherein a program in the software code performs the steps of the method described in the first aspect of the present invention.

[0011] The above technical solution has the following advantages or beneficial effects: This invention aims to minimize the operating cost of microgrids throughout the entire lifecycle of energy storage. It establishes a long-cycle optimization model that considers energy storage lifetime losses and the transfer of state of charge across operating days. Addressing the problem of the model's large scale and difficulty in direct solution, this invention uses the operating day as the decision-making stage, connecting adjacent operating days with the accumulation of energy storage lifetime losses and the transfer of state of charge. This decomposes the full-cycle operation optimization into stage optimizations for each operating day, forming a Markov decision process recursively applied to the operating day. Thus, the original problem is transformed into a multi-stage decision problem solvable on an operating day basis.

[0012] This invention approximates the value function using the cumulative lifetime loss state and state of charge (SBC) of energy storage, respectively. It simplifies the value function calculation using a piecewise linear function and incorporates it into daily stage optimization, enabling daily stage decisions to consider both long-term utilization of energy storage lifetime and cross-operational SBC adjustment. Furthermore, it designs a value function slope sampling and iterative update method and improves convergence efficiency while maintaining the approximate accuracy of the value function through a multi-precision training strategy that combines merging piecewise steps with piecewise fine-tuning updates. Attached Figure Description

[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0014] Figure 1 This is a flowchart of the microgrid energy storage full life cycle optimization operation method in a specific embodiment of the present invention; Figure 2 This is a flowchart illustrating the training and full-cycle scheduling using approximate dynamic programming theory in this invention. Figure 3 For node microgrid systems; Figure 4 This is a comparison chart of the cumulative lifetime loss of different schemes during experimental verification; Figure 5 A comparison chart of the daily marginal value of energy storage lifetime loss for different schemes during experimental verification; Figure 6 A comparison chart of monthly lifespan loss and monthly cost savings for different schemes during experimental verification; Figure 7 This refers to the change in net energy storage capacity in August for Scheme 6; Figure 8 The August load and electricity price changes for Scheme 6; Figure 9 The energy storage charging and discharging operation of Scheme 5 is shown in the data for 230-231 days. Figure 10 The energy storage charging and discharging operation of Scheme 6 is shown in the data for 230-231 days. Figure 11 The training curves for schemes 5 and 6 are shown. Detailed Implementation

[0015] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0016] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the invention. The terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] In this embodiment of the invention, "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of this invention, "multiple" refers to two or more.

[0018] Furthermore, to facilitate a clear description of the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0019] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0020] Example 1 The key to optimizing the entire lifecycle of microgrid energy storage lies in coordinating the frequency of energy storage usage across different operating days and rationally adjusting the state of charge (SOC). On one hand, load, electricity price, and renewable energy output differ across operating days, necessitating a rational allocation of energy storage lifecycle usage based on the actual operational value of each stage, ensuring that energy storage is used more during high-value operating periods. On the other hand, the SOC at the end of the energy storage day carries over to the next operating day, and its changes affect subsequent charging and discharging capabilities and operational benefits. Therefore, optimizing the entire lifecycle of energy storage is essentially a long-term scheduling problem that coordinates lifecycle allocation with SOC adjustment across operating days. However, the service life of energy storage is typically measured in years, and energy storage lifecycle depletion and SOC state exhibit strong coupling characteristics across different operating days. Directly constructing and solving a scheduling model covering the entire lifecycle results in a massive model with high solution complexity.

[0021] To solve the above problems, such as Figure 1 As shown in the figure, this embodiment provides a method for optimizing the operation of microgrid energy storage throughout its entire life cycle, including: S101: Obtain microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total life cycle operating cost; S102: Reconstruct the microgrid scheduling model into a Markov decision process according to the operating day; S103: Construct the cumulative lifetime loss state value function and the charge state value function for each operating day, and perform piecewise linear approximation on both functions to determine the initial slope of each segment; construct the optimization objective based on the operating cost, cumulative lifetime loss state value function, and charge state value function of the current operating day according to the order of operating days, and solve the scheduling decision for the current operating day; update the initial state of the next operating day based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round; S104: Based on the scheduling decisions for each operating day of this round, calculate the slope sampling values ​​of the two value functions respectively, and update the slope of each segment accordingly; repeat the scheduling decisions for all operating days and subsequent slope updates until the two value functions meet the convergence conditions; adopt a multi-precision accelerated training strategy during the iterative update of the two value functions; in the initial update stage, divide multiple consecutive target segments into several sub-intervals, so that multiple segments within the same sub-interval share the same slope, and update the shared slope of the sub-intervals according to the scheduling results of the operating days; after obtaining the overall trend of the value functions, restore the segments within each sub-interval to independent segments, and use the convergence slope of the corresponding sub-interval as the initial slope of each independent segment, and perform slope sampling and updating respectively; S105: Using the two converged value functions, the scheduling decision is executed again in the order of the operating days to obtain the final scheduling decision for each operating day.

[0022] This invention first uses the operating day as the decision-making stage, decomposing and reconstructing the long-term microgrid scheduling model into a Markov Decision Process (MDP). The impact of the current operating day's scheduling result on subsequent operations is described through state transition relationships, transforming the entire lifecycle optimization into a multi-stage decision problem recursively applied to operating days. Furthermore, based on Approximate Dynamic Programming (ADP) theory, approximate value functions are constructed by selecting energy storage lifetime states and State of Charge (SOC) states that can characterize the coupling relationship across operating days. The stage scheduling model is solved sequentially according to the operating days. The slope of the value function is sampled and iteratively updated based on the optimization results of each day until the value function converges, thereby obtaining an approximate value function that can assist in daily optimized scheduling while considering both current operating costs and subsequent operational impacts. This application avoids excessive consumption of finite lifetime energy on low-value operating days, crowding out subsequent more valuable usage opportunities, prioritizing the use of finite lifetime energy on higher-yield operating days, and determining how much energy should be reserved today for future use.

[0023] Step S101 is as follows: S101.1: Obtain microgrid operating parameters and construct a microgrid scheduling model with the objective function of minimizing the microgrid operating cost throughout the entire life cycle of energy storage.

[0024] The objective function of the microgrid dispatch model includes the gas turbine cost and the cost of purchasing and selling electricity from the external grid, i.e.: (1); (2); In the formula: F is the net operating cost of the microgrid over the entire life cycle of energy storage (excluding energy storage costs); The operating cost of the microgrid on day d; The output power of the gas turbine (GT) at time t on day d; The external power grid price at time t on day d; The power exchanged between the microgrid and the external power grid at time t on day d; , and For GT cost parameters; For external power grid collection; This refers to the total number of operating days throughout the entire lifecycle of the energy storage system. This refers to the number of time steps within a single day. In this embodiment, it is divided into hours. .

[0025] This embodiment aims to minimize the total operating cost of the microgrid throughout the entire life cycle of energy storage. It coordinates the charging and discharging strategies of energy storage at different operating stages over a long time scale to improve the overall efficiency of energy storage throughout its entire life cycle.

[0026] S101.2: Determine the constraints of the microgrid dispatch model. The constraints include unit ramping constraints, energy storage operation constraints, energy storage lifetime loss constraints, power flow constraints, and security constraints.

[0027] S101.2.1: Constructing unit ramp-up constraints: (3); In the formula, Day d The output power of the gas turbine at any given time; and These represent the upward and downward ramp power of unit g, respectively; It is a collection of gas turbines.

[0028] S101.2.2: Constructing energy storage operation constraints: (4); (5); (6); (7); In the formula: The energy storage capacity at time t on day d; For energy storage on day d Time capacity, and These represent the maximum and minimum values ​​of the energy storage capacity, respectively. and These are the active charging power and discharge power of energy storage at time t on day d, respectively. and These are the charging and discharging efficiencies of energy storage, respectively. and These are the maximum active charging power and the maximum active discharging power of the energy storage, respectively.

[0029] S101.2.3: Constructing energy storage lifetime loss constraints: (8); (9); In the formula: This is the energy storage lifespan loss coefficient; For day d, from time t to Lifespan that is constantly diminishing; This refers to the total usable lifespan of energy storage throughout its entire lifecycle.

[0030] S101.2.4: Constructing power flow constraints: (10); (11); In the formula: and These represent the active and reactive power injected into node i by the generator unit or the external power grid k at time t on day d, respectively. and Let represent the active charging and discharging power of energy stored at node i at time t on day d, respectively; and Let represent the reactive charging and discharging power of energy stored at node i at time t on day d, respectively; and These represent the active and reactive loads at node i, respectively; and These are the conductance and susceptance on branch ij, respectively. The susceptance on branch ij without parallel capacitors; and These are the voltage magnitude and phase angle at node j, respectively; , , and These are respectively a set of wind turbine units, a set of photovoltaic units, a set of microgrid nodes, and a set of microgrid branches.

[0031] S101.2.5: Constructing security constraints: (12); (13); (14); (15); (16); (17); (18); (19); (20); (twenty one); (twenty two); (twenty three); In the formula: and These represent the maximum and minimum active power values ​​for the GT unit, respectively. and Let w represent the actual power generation of wind turbine w and photovoltaic unit s on day d at time t, respectively. and Let w represent the available power of wind turbine w and photovoltaic unit s on day d at time t, respectively. , , , The reactive power generated by GT unit g, external power grid e, wind turbine w, and photovoltaic unit s on day d at time t. , , , , and These represent the maximum reactive power values ​​for GT unit g, external power grid e, wind turbine w, photovoltaic unit s, energy storage charging, and energy storage discharging, respectively. and These are the maximum and minimum voltage amplitudes at node i, respectively; and These are the maximum and minimum voltage phase angles at node i, respectively.

[0032] Step S102 is as follows: To address the difficulty of solving long-term optimization problems in energy storage, this embodiment uses the operating day as the decision-making stage, decomposing and reconstructing the original long-term scheduling problem into an MDP, thereby forming a multi-stage decision-making model recursively based on the operating day. The following sections define the inter-operating day transition constraints, state variables, decision variables, and state transition equations.

[0033] S102.1: Construct and run daytime transfer constraints.

[0034] If the long-cycle microgrid scheduling model is divided into multi-stage decision-making based on the operating day, in order to maintain the equivalence between the reconfiguration model and the original model, it is necessary to establish a connection between the beginning and end states at the boundary of the operating day, that is, the first and last states are connected. d The end-of-day state of the day as the first d The initial state after +1 day.

[0035] (twenty four); (25); (26); (27); (28); In the formula: and These represent the cumulative energy storage lifetime loss at the beginning and end of day d, respectively. and For the first d Energy storage capacity at the beginning and end of the day; For the first d Heaven t Time's up Changes in energy storage capacity at any given time.

[0036] S102.2: Define state variables.

[0037] Since energy storage lifespan loss and state of charge change with the energy storage behavior of the current operating day, and further affect the scheduling mode of subsequent operating days, the [number]th [day] is selected. d At the start of the day, the accumulated lifetime loss and state of charge of the energy storage are used as state variables. Simultaneously, to describe the impact of the operating environment on scheduling decisions on the day, the [missing information] is used... d Information on wind power, photovoltaic power, load, and electricity price is incorporated into the state variables.

[0038] (29); In the formula: and They are respectively The matrix formed and The matrix formed represents the first... d Available power generation capacity of wind turbines and photovoltaic units at different times of the day; for The T-dimensional vector represents the first... d Electricity prices for different times of day; and They are respectively The matrix formed and The matrix formed represents the first... d Active and reactive loads at each time period and at each node throughout the day.

[0039] S102.3: Define decision variables.

[0040] Decision variables are used to describe the scheduling behavior of the microgrid during the current operating day, including variables such as gas turbine output, power purchased and sold from the external grid, energy storage charging and discharging power, and wind and solar power output.

[0041] (30); In the formula: and They are respectively The matrix formed and The matrix formed represents the first... d Energy storage charging and discharging power at different times of the day; for The T-dimensional vector represents the first... d Energy storage lifespan loss at different times of the day; and They are respectively The matrix formed and The matrices formed represent the first, second, and third parts of the matrix. d Actual power generation of wind turbines and photovoltaic units at different times of the day; for The T-dimensional vector represents the first... d The trading power between the microgrid and the external power grid at different times of the day; and They are respectively The matrix formed and The matrices formed represent the first, second, and third parts of the matrix. d Voltage amplitude and phase angle at each node during different time periods of the day.

[0042] S102.4: Construct the state transition equation.

[0043] Equations (24) to (28) describe the energy transfer process between operating days and the initial and final state relationships between adjacent operating days. Based on the above constraints, the first... d The cumulative lifespan loss and state of charge changes of energy storage resulting from daily dispatch decisions will be updated to determine the [number].d The initial state is +1 day. Therefore, the state transition equation in the MDP can be expressed as: (31); (32); in, For the first The first item in the daily state variables; The first term in the state variable on day d (i.e., in equation (29)) ); The third term in the action variables on day d (i.e., in equation (30)) ); For the first The second item in the daily state variables; It is the second item in the state variables on day d; This is the first item in the action variables on day d. It is the second item in the action variables on day d.

[0044] Based on the above MDP reconfiguration, the full lifecycle scheduling of microgrid energy storage can be represented as a decision-making process recursively calculated on an operational day basis. d At the beginning of the day, the state transition equation is used to obtain the state from the first day. The energy storage status at the end of the day was obtained. d The beginning of the sun and The status is obtained by combining information such as wind power, solar power, load, and electricity price on that day. Then solve for the first... d The daily stage scheduling problem yields the decision variables for that day. .implement Later obtained and Information for updating d +1 day status The process continues into the next day, continuing until the microgrid operation decision for the entire scheduling cycle is completed. This transforms the original large-scale, full-lifecycle optimization problem into multiple daily-stage optimization problems, while preserving the coupling relationship between cumulative energy storage lifetime loss and state of charge across different operating days.

[0045] Step S103 is as follows: The cumulative lifetime loss state-value function and the charge state-value function of energy storage are defined and approximated.

[0046] After MDP reconstruction, the multi-stage decision problem recursively based on the operating day can be solved sequentially for the optimal strategy under each operating day state based on the Bellman optimality principle. Its Bellman recursive form is: (33); In the formula: State on day d Execute the daily scheduling decision The obtained stage operating costs; State on day d The minimum cumulative cost from the current running period to the end of the scheduling cycle, representing the state on day d. The corresponding future value function.

[0047] No. d The optimal decision for a given day can be expressed as: (34); If a value function is constructed using complete state information, a complex state-value mapping needs to be learned. As the state dimension increases, the calculation and training of the value function will face a significant computational burden. Since the value function in MDP is used to characterize the impact of the current stage state on subsequent operations, the state variables used to construct the value function should be selected from those that participate in the inter-day coupling constraints and can transmit the impact of the current stage decision on subsequent stages. As can be seen from equations (24) to (28) and (31) to (32), both the cumulative lifetime loss of energy storage and the state of charge participate in the state transition across operating days, respectively determining the subsequent available lifetime space and energy level, and are key state variables connecting adjacent operating days. Therefore, this embodiment selects the initial cumulative lifetime loss. and initial energy storage capacity A dimension reduction approximation of the value function is performed, namely: (35); Then equation (33) can be further expressed as: (36); Among them, the cumulative lifetime loss state value function This value function is used to characterize the impact of current lifetime consumption on the operating cost of the remaining scheduling cycle. Since the total lifetime of energy storage is finite and irrecoverable, this function avoids over-utilizing energy storage in pursuit of only daily gains and guides the limited lifetime to prioritize operating days with higher operational value. State of Charge Value Function This function is used to characterize the impact of end-of-day energy state on subsequent operating costs. It can adjust the end-of-day state of charge based on subsequent load, electricity price, and renewable energy output, reserving a suitable energy state for future operation. The former focuses on the long-term allocation of lifetime resources, while the latter focuses on the daytime coordination of energy state. Together, they achieve synergistic optimization of energy storage lifetime allocation and energy state regulation.

[0048] because and Since all variables are continuous and the energy storage lifecycle is relatively long, representing the value function in tabular form would require setting a large number of discrete state points, leading to a significant training burden. Therefore, this embodiment constructs piecewise linear approximate value functions for cumulative lifetime loss and state of charge for each operating day. The slope of these functions, varying with the operating day and state segment, characterizes the impact of different energy storage states on subsequent operating costs. When the energy storage is in a given lifetime and state of charge, the approximate value function is calculated based on the slope of each segment and the state value in the corresponding segment. This approximate value function, along with the daily operating cost, constitutes the stage optimization objective, thereby guiding the daily scheduling decision. The cumulative lifetime loss interval is divided into... The state of charge interval is divided into segments. If there are multiple segments, then equation (33) can be further rewritten as: (37); (38); (39); (40); (41); In the formula: and They are respectively The slopes of the cumulative lifetime loss state-value function and the charge state-value function on their respective segments; and These represent the values ​​of the cumulative lifetime loss state-value function and the charge state-value function on the corresponding segments, respectively.

[0049] Therefore, equation (34) can be further expressed as: (42); Through the above processing, the problem of solving the value function that depends on the high-dimensional state throughout the entire life cycle is transformed into a dimension-reduced approximation that only considers the cumulative life loss and state of charge of energy storage, and is further expressed as the problem of solving the slope of each segment.

[0050] Step S104 is as follows: The slopes of the cumulative lifetime loss state-value function and the charge state-value function of energy storage are sampled and updated. A piecewise linear function is used to approximate the value function. The key to value function training lies in obtaining the slope of each piecewise segment. To obtain a good value function slope, iterative training of the ADP (Active Dynamic Programming) is required.

[0051] In the ADP training n In the nth iteration, based on the 1st... n- The piecewise slope obtained in the first iteration is used to solve for the th iteration. d The day-stage optimization problem, namely: (43); In the formula: and The first n In the -1st iteration, the th The slopes of the cumulative lifetime loss function and the SOC state value function on the corresponding segments.

[0052] Solving equation (43) yields the first... n The second iteration d The scheduling decision of the day And the stage-optimal objective value, i.e., the approximate value function. To update the number d The piecewise slope corresponding to the state of charge requires calculating the marginal changes of the value function with respect to cumulative lifetime loss and state of charge. Since the piecewise linear function may not have a strict derivative at the breakpoints, this embodiment uses a difference method to calculate the slope sample value. Therefore, the value function slope sample value can be expressed as: (44); (45); In the formula: and These represent the slope sampling estimates of the piecewise linear functions of cumulative lifetime loss and state of charge of energy storage, respectively, in the corresponding segments.

[0053] After obtaining the slope sample value, based on the slope of the value function in the previous iteration and the slope sample value in the current iteration, the slope of the current segment of the value function is corrected using a weighted update method, that is: (46); in, To update the step size, and Update step size This step size is mainly used to adjust the weight between the historical slope and the current sampling slope, and its value is usually selected empirically based on the training convergence performance. A larger update step size can increase the weight of the current sampling information and speed up the value function update, but may increase the fluctuations during the iteration process; a smaller update step size can improve training stability, but the convergence speed is relatively slow. Therefore, the update step size can be determined by comparing the pre-training results under different step sizes within the range of (0,1) and comprehensively considering the convergence speed and the stability after convergence. In this embodiment, 0.9 is specifically selected.

[0054] In energy storage lifecycle training, a single complete iteration requires sequentially optimizing all operating days' stages and continuously updating cumulative lifetime loss and state of charge. Training the slope directly across all segments not only results in a large computational load per iteration, but also makes it difficult to fully access some lifetime loss and state of charge segments in the early stages of training, leading to slow slope updates for the value function. To adapt to the value function training requirements under a recursive approach based on operating days throughout the entire lifecycle, this embodiment proposes a multi-precision accelerated training method. The specific process is as follows: Initial value calculation: First determine the total number of target segments. ,Will Each segment is divided into the following sections in sequence: consecutive sub-intervals ( ). same All segments within a subinterval maintain the same slope value, meaning an equality constraint is added to the original segments: (47); In the formula: In the nth iteration, the th... m The slope values ​​of each segment; For the first k The uniform slope corresponding to each sub-interval.

[0055] When the system status is accessed any segment within m The slope sampling value is calculated according to equation (44) or (45). Then used to update the uniform slope of the subintervals. : (48); In this way, the number of slope parameters to be trained is reduced from... The number was reduced to This allows for a rapid preliminary estimate of the slope of each merged interval with fewer iterations. Let the convergence slope of each sub-interval at the time of convergence in this stage be... .

[0056] Perform segment-by-segment updates to obtain the overall trend of the value function. Each segment resumes its independent update state. At this point, the uniform slope obtained in the first stage is assigned to each small segment within the corresponding interval as the initial slope for subsequent precise optimization: (49); In the formula: Let be the initial value for the iteration of the m-th segment.

[0057] Based on this, according to equation (44) or (45) The slope of each segment is sampled and updated independently until the training and parameter updates of all segments are completed.

[0058] This embodiment obtains the cumulative lifetime loss value function and the state of charge value function through iterative training, and uses the trained value functions to assist in daily scheduling. The specific process is as follows: Figure 2 As shown: Step 1: Generate training scenarios. Based on historical load, wind power, photovoltaic output, and electricity price data, generate typical operating scenarios.

[0059] Step 2: Set the current training phase. Determine whether you are currently in the initial value calculation phase or the segmented update phase.

[0060] Step 3: Initialize training parameters. Initialize the slope of the value function, and set the total number of training iterations and the number of segments. and Update step size etc.; at the start of each training iteration, the state variables of the first day are restored to their initial values.

[0061] Step 4: Run daily phase optimization. Based on the current status. Solve the daily optimization scheduling model to obtain the stage optimization results used for slope sampling and value function updating. .

[0062] Step 5: Slope Sampling and Update. Based on the optimization results of day d. The slope sample value is calculated using a differential method. In the initial value calculation stage, the slope of the merged interval is corrected; in the segment-by-segment update stage, the slope of the corresponding segment is corrected.

[0063] Step 6: State Recursion. Update the state for the next operating day based on the cumulative lifetime loss value on day d and the state of charge at the end of the day. .

[0064] Step 7: Determine if all runtime iterations have been completed. If not, repeat steps 4 to 6 until all runtimes of the current iteration have been completed; if yes, proceed to the next step. Step 8: Determine if convergence has occurred. If not, repeat steps 3 to 7 until iterative convergence occurs; if yes, proceed to the next step.

[0065] Step 9: Determine if all running phase iterations have been completed. If not, repeat steps 2 to 8 until all running phases are completed; if yes, proceed to the next step.

[0066] Step 10: Output the value function results. Save the final cumulative lifetime loss value function slope and the state of charge value function slope to obtain the trained value function.

[0067] Step 11: Assisting in the actual scheduling of energy storage throughout its entire lifecycle. During the actual operation of energy storage, the value function obtained from training is introduced as a future value term into the daily optimization, and the entire lifecycle scheduling is completed sequentially according to the daily operation.

[0068] Experimental verification A 33-node microgrid test system was constructed based on the IEEE 33-node standard system to verify the effectiveness of the proposed method. Figure 3 As shown in Table 3, the energy storage lifespan is set to 20 years, with a continuous one-year scheduling cycle, and the upper limit of energy storage lifespan loss within the cycle is set to 5% of the total lifespan. The parameters of the gas turbine, wind turbine, and photovoltaic unit are shown in Tables 1 and 2, and the parameters of the energy storage device are shown in Table 3.

[0069] Table 1 Parameter Configuration of Wind Power and Solar Power Units

[0070] Table 2 Gas Turbine Parameter Configuration

[0071] Table 3 Energy Storage Parameter Configuration

[0072] The annual load and wind / solar output time-series curves and electricity price data are based on data from https: / / transparency.entsoe.eu / dashboard / show. Energy storage costs are set at 0.35 yuan / Wh. The simulation was performed using MATLAB with the GUROBI solver. The computer used for the simulation had a Windows 11 system, an Intel Core i7-14650HX CPU with a base clock speed of 2.20GHz, and 32GB of RAM.

[0073] To verify the effectiveness of the proposed method in improving the operational efficiency of energy storage, this embodiment sets up the following six comparison schemes: Option 1: Microgrid operation without energy storage; Option 2: Microgrid operation with energy storage but without considering energy storage lifespan loss (only minimizing daily operating costs); Option 3: Microgrid operation considering energy storage lifetime loss (the product of the current cycle lifetime loss ratio and the energy storage replacement cost is included in the objective function). Option 4: Microgrid operation considering energy storage lifetime loss (lifetime sharing strategy, setting a daily lifetime loss limit). Option 5: Only adopt the energy storage lifetime allocation strategy (considering the cumulative lifetime loss state, but not the daytime transfer of state of charge operation). Solution 6: The energy storage lifetime and state of charge optimization strategy proposed in this invention (considering both cumulative lifetime loss state and daytime transfer of state of charge operation).

[0074] Operating cost results analysis: To verify the effectiveness of the microgrid dispatch model proposed in this embodiment, which considers the cumulative lifetime loss of energy storage and the daily changes in energy storage capacity, the operating results under the six optimization schemes are shown in Table 4. Here, energy storage benefit is defined as the total operating cost saving when the energy storage system is decommissioned compared to the system without energy storage; net rate of return is the ratio of energy storage benefit minus energy storage investment cost to the energy storage investment cost.

[0075] Table 4 Comparison of Operating Costs and Energy Storage Benefits of Six Schemes

[0076] As shown in Table 4, compared to Schemes 1 to 5, Scheme 6 reduces net operating costs by 25.07%, 20.67%, 9.86%, 8.91%, and 5.24%, respectively, while increasing energy storage efficiency by 325.54%, 40.17%, 33.31%, and 19.60%, respectively. Scheme 6 has the lowest net operating cost and the highest energy storage efficiency among all schemes.

[0077] Option 2 reduces operating costs by introducing energy storage, but it fails to allocate lifetime resources rationally, leading to excessively rapid consumption of energy storage in the early stages, and the total lifecycle revenue may not even cover the investment cost. Options 3 and 4 employ lifetime attrition cost conversion and lifetime sharing constraints, respectively, which can curb excessive energy storage use to some extent, but they fail to rationally determine the daily energy storage usage intensity based on differences in different operating days, leaving room for improvement in energy storage utilization efficiency. Option 5 uses the ADP value function to characterize the impact of remaining lifetime on subsequent operating costs, allowing for adjustments to daily lifetime usage based on changing operating conditions, ensuring that the limited lifetime is used more for higher-value operating days, thereby improving the overall lifecycle utilization efficiency of energy storage. Option 6, building on Option 5, further considers the transfer of SOC between adjacent operating days, allowing for adjustments to the initial daily SOC based on subsequent operating conditions, thus further reducing operating costs.

[0078] Lifetime allocation results analysis: To reveal the advantages of Schemes 5 and 6 in terms of lifetime allocation, the lifetime loss process and benefits of schemes containing energy storage are analyzed. For example... Figure 4 and Figure 5 As shown, the energy storage lifetime loss and daily marginal value of energy storage lifetime under different schemes are presented on a daily scale. Figure 6 The changes in energy storage lifetime loss and monthly cost savings under different schemes are shown on a monthly scale. Figure 5The “marginal value per energy storage lifetime” refers to the cost savings achieved by consuming one unit of energy storage lifetime on a given operating day.

[0079] Depend on Figure 4 It is evident that Schemes 5 and 6 exhibit slower initial lifespan degradation, but their lifespan depletes significantly in the later stages. Combined with... Figure 5 It can be seen that the periods of increased lifespan investment in Schemes 5 and 6 are basically consistent with the periods when the marginal value of days with higher lifespan losses is concentrated. This indicates that the method proposed in this invention incorporates subsequent operating revenue information into the decision-making process on the same day through a value function, and allocates more of the limited lifespan to operating days with higher unit lifespan revenue. This reflects the forward-looking nature of the proposed strategy in the allocation of energy storage lifespan.

[0080] Figure 6 The above-mentioned lifetime allocation mechanism was further verified on a monthly scale. Scheme 3 consumed a significant portion of its lifetime in the early stages and exhausted it after the 9th month, thus missing the period of greater cost savings and higher energy storage operational value in months 10-12. While Scheme 4 employs a fixed, evenly distributed strategy and cannot proactively adjust lifetime investment based on monthly operational value differences, it ensures that energy storage can still participate in operation later on, thus achieving higher annual returns than Scheme 3. In contrast, Schemes 5 and 6 increase lifetime utilization when operational cost savings are higher in months 10-12, indicating that they can adjust lifetime investment based on operational value at different times, thereby improving energy storage utilization efficiency. Further comparison shows that the overall monthly lifetime loss is similar for both schemes, but Scheme 6 achieves higher operational cost savings, indicating that daytime transfer during SOC operation can improve energy state, resulting in higher operational returns for similar lifetime investments.

[0081] Analysis of the daytime SOC regulation mechanism: To further reveal the underlying mechanism of SOC optimization for energy storage in Scheme 6, the end-of-day state of charge variation law of energy storage in Scheme 6 is analyzed.

[0082] Depend on Figures 7-8 It is evident that when the daily electricity price is high and the electricity demand is large on a certain operating day, Scheme 6 will appropriately increase the end-of-day SOC on the previous operating day to reserve available electricity for the high-price period, thereby reducing the microgrid's need to purchase electricity at high prices. It is also clear that the proposed method can adjust the daily SOC by combining the load and price differences between adjacent operating days, ensuring that the charging decision on the previous day also considers the discharge demand in subsequent periods, thus improving the energy regulation benefits of energy storage across operating days.

[0083] like Figures 9-10As shown, the above adjustment process is further illustrated using day 230-231 as an example. In Scheme 5, the State of Charge (SOC) needs to return to its initial value at the end of the day, and charging and discharging are based on the peak-valley price difference for both days, mainly achieving intraday energy regulation. Scheme 6, on the other hand, retains some electricity after charging during the low-price period on day 230 and uses it for discharging during the higher-price period on day 231. The marginal lifetime value of energy storage in Schemes 5 and 6 over 48 hours is calculated to be RMB 618,600.74 and RMB 1,611,560.69, respectively. It is evident that considering the intraday transfer of SOC provides more flexibility for energy storage charging and discharging decisions, allowing SOC regulation to no longer be limited to a single operating day, but to take into account the discharge demand of subsequent periods in the current charging decision, thereby enhancing the energy storage's ability to regulate load and electricity price changes over time.

[0084] Analysis of Multi-Precision Accelerated Iteration Results: To verify the effectiveness of the proposed ADP-accelerated iteration method, the training processes of schemes 5 and 6 were analyzed, and their training curves are shown below. Figure 11 As shown.

[0085] In this example, the total number of segments is set to 30. In the first 100 iterations, segments 1-10, 11-20, and 21-30 are merged for calculation. Using 202993.85 yuan and 192339.17 yuan from Table 4 as stable reference values ​​for Scheme 5 and Scheme 6, respectively, Scheme 5 reached a stable value within the 0.1% error range for the first time in 679 and 299 iterations, respectively, while Scheme 6 reached 721 and 293 iterations, respectively. This is because merging segments in the initial value calculation stage reduces the parameters to be updated, allowing for a faster acquisition of a reasonable initial value for the slope of the value function. Subsequently, the 30 segments are trained segment by segment, eventually converging to a stable level consistent with the unaccelerated training. The results show that the multi-precision segmented accelerated training method proposed in this invention can significantly improve model training efficiency while ensuring consistent final convergence results.

[0086] Based on the above numerical examples, the following main conclusions can be drawn: The method proposed in this invention can achieve coordinated optimization of energy storage lifetime and state of energy (SOC) over long periods. The system can adaptively allocate energy storage lifetime resources and flexibly adjust the initial daily SOC level according to load levels, electricity price fluctuations, and differences in renewable energy output on different operating days throughout its entire lifecycle. This effectively avoids excessive short-term use of energy storage, reduces the long-term operating cost of the microgrid, and significantly improves the overall economic efficiency of energy storage throughout its entire lifecycle. The approximate dynamic programming algorithm introduces a value function to approximate the impact of the current decision on future long-term operating benefits, and combines this with a multi-precision accelerated training method to improve the learning efficiency of the value function, thereby effectively solving the problem of solving long-term, cross-period coupling in the whole lifecycle scheduling of microgrids.

[0087] Example 2 This embodiment provides a microgrid energy storage full lifecycle optimized operation system, including: The model building module is configured to: acquire microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total life cycle operating cost; The model reconstruction module is configured to reconstruct the microgrid scheduling model into a Markov decision process according to the operating day. The daily decision-making module is configured to: construct the energy storage cumulative lifetime loss state value function and the charge state value function for each operating day, and perform piecewise linear approximation on both functions to determine the initial slope of each segment; construct an optimization objective based on the operating cost, energy storage cumulative lifetime loss state value function, and charge state value function of the current operating day according to the operating day sequence, and solve the scheduling decision for the current operating day; update the initial state of the next operating day based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round; The iterative training module is configured to: calculate the slope sampling values ​​of the two value functions according to the scheduling decisions for each running day of the round, and update the slope of each segment accordingly; repeat the scheduling decisions and subsequent slope updates for all running days until the two value functions meet the convergence conditions; adopt a multi-precision accelerated training strategy during the iterative update of the two value functions; in the initial update phase, divide multiple consecutive target segments into several sub-intervals, so that multiple segments within the same sub-interval share the same slope, and update the shared slope of the sub-intervals according to the scheduling results of the running days; after obtaining the overall trend of the value functions, restore the segments within each sub-interval to independent segments, and use the convergence slope of the corresponding sub-interval as the initial slope of each independent segment, and perform slope sampling and updating respectively; The final scheduling module is configured to: use the two converged value functions to execute the scheduling decision again in the order of the running days to obtain the final scheduling decision for each running day.

[0088] It should be noted that the model building module, model reconstruction module, daily decision-making module, iterative training module, and final scheduling module mentioned above correspond to steps S101 to S105 in Embodiment 1. The examples and application scenarios implemented by these modules and their corresponding steps are the same, but they are not limited to the content disclosed in Embodiment 1. It should be noted that these modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0089] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0090] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0091] Example 3 This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the method described in Embodiment 1.

[0092] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0093] Memory may include read-only memory and random access memory, and provides instructions and data to the processor. A portion of memory may also include non-volatile random access memory. For example, memory may also store information about the device type.

[0094] In the implementation process, each step of the above method can be completed by the integrated logic circuits in the processor hardware or by software instructions.

[0095] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0096] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0097] Example 4 Embodiment 4 of the present invention provides a computer-readable storage medium.

[0098] A computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the steps of the method as described in Embodiment 1 of the present invention.

[0099] The detailed steps are the same as those provided in Example 1, and will not be repeated here.

[0100] Example 5 Embodiment 5 of the present invention provides a computer program product.

[0101] A computer program product includes software code, wherein the program in the software code performs the steps described in Embodiment 1 of the present invention.

[0102] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for optimizing the operation of microgrid energy storage throughout its entire lifecycle, characterized in that, include: Obtain microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total lifecycle operating cost; The microgrid scheduling model is reconstructed into a Markov decision process according to the operating day; The cumulative lifetime loss state value function and the charge state value function for each operating day are constructed respectively, and piecewise linear approximations are performed on the two value functions to determine the initial slope of each segment. According to the order of operating days, an optimization objective is constructed based on the operating cost, the cumulative lifetime loss state value function, and the charge state value function of the current operating day, and the scheduling decision for the current operating day is solved. The initial state of the next operating day is updated based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round. Based on the scheduling decisions for each operating day of this round, the slope sampling values ​​of the two value functions are calculated respectively, and the slope of each segment is updated accordingly; the scheduling decisions for all operating days and subsequent slope updates are repeated until the two value functions meet the convergence condition; a multi-precision accelerated training strategy is adopted in the iterative update process of the two value functions; in the initial update stage, multiple consecutive target segments are divided into several sub-intervals, so that multiple segments in the same sub-interval share the same slope, and the shared slope of the sub-intervals is updated according to the scheduling results of the operating days; after obtaining the overall trend of the value functions, the segments in each sub-interval are restored to independent segments, and the convergence slope of the corresponding sub-interval is used as the initial slope of each independent segment, and slope sampling and updating are performed respectively; Using the two converged value functions, the scheduling decision is executed again in the order of the operating days to obtain the final scheduling decision for each operating day.

2. The microgrid energy storage full life cycle optimized operation method as described in claim 1, characterized in that, The constraints of the microgrid dispatch model include: unit ramping constraints, energy storage operation constraints, energy storage lifetime loss constraints, power flow constraints, and security constraints.

3. The microgrid energy storage full life cycle optimized operation method as described in claim 1, characterized in that, The state variables of the Markov decision process include: the cumulative lifetime loss and state of charge of the energy storage at the beginning of the current operating day, as well as the wind power output, photovoltaic power output, load and electricity price information of the current operating day; the decision variables of the Markov decision process include: gas turbine output, power purchased and sold from the external grid, energy storage charging and discharging power and wind and solar power output; the state transition relationship of the Markov decision process is: the cumulative lifetime loss and state of charge of the energy storage corresponding to the energy storage scheduling result of the current operating day are used as the initial state of the next operating day.

4. The microgrid energy storage full life cycle optimized operation method as described in claim 1, characterized in that, The optimization objective is constructed based on the operating cost, cumulative lifetime loss state value function, and charge state value function of the current operating day. Specifically, the sum of the operating cost of the current operating day and the cumulative lifetime loss state value function and charge state value function is used as the optimization objective of the current operating day. The energy storage scheduling decision for the current operating day is determined by minimizing the result of the optimization objective. The value of the state value function is calculated based on the segment where the current state is located and the corresponding slope.

5. The microgrid energy storage full life cycle optimized operation method as described in claim 1, characterized in that, The slope sampling values ​​of the two value functions are calculated separately, and the slope of each segment is updated accordingly. Specifically, the marginal changes of the value function with respect to the cumulative lifetime loss of energy storage and the state of charge are calculated using the difference method to obtain the slope sampling estimate value on each segment. Based on the slope of the value function in the previous iteration and the slope sampling estimate value in the current iteration, the slope of the current segment is corrected using a weighted update method to complete the iterative training of the value function.

6. The microgrid energy storage full life cycle optimized operation method as described in claim 2, characterized in that, The energy storage lifespan loss constraint is specifically defined as follows: the sum of lifespan losses for each operating day and time period within the entire lifespan of the energy storage is equal to the total usable lifespan of the energy storage.

7. A microgrid energy storage full life cycle optimized operation system, characterized in that, include: The model building module is configured to: acquire microgrid operating parameters and construct a microgrid scheduling model with the goal of minimizing the total life cycle operating cost; The model reconstruction module is configured to reconstruct the microgrid scheduling model into a Markov decision process according to the operating day. The daily decision-making module is configured to: construct the energy storage cumulative lifetime loss state value function and the charge state value function for each operating day, and perform piecewise linear approximation on both functions to determine the initial slope of each segment; construct an optimization objective based on the operating cost, energy storage cumulative lifetime loss state value function, and charge state value function of the current operating day according to the operating day sequence, and solve the scheduling decision for the current operating day; update the initial state of the next operating day based on the scheduling decision of the current operating day to complete the scheduling decision for all operating days in one round; The iterative training module is configured to: calculate the slope sampling values ​​of the two value functions according to the scheduling decisions for each running day of the round, and update the slope of each segment accordingly; repeat the scheduling decisions and subsequent slope updates for all running days until the two value functions meet the convergence conditions; adopt a multi-precision accelerated training strategy during the iterative update of the two value functions; in the initial update phase, divide multiple consecutive target segments into several sub-intervals, so that multiple segments within the same sub-interval share the same slope, and update the shared slope of the sub-intervals according to the scheduling results of the running days; after obtaining the overall trend of the value functions, restore the segments within each sub-interval to independent segments, and use the convergence slope of the corresponding sub-interval as the initial slope of each independent segment, and perform slope sampling and updating respectively; The final scheduling module is configured to: use the two converged value functions to execute the scheduling decision again in the order of the running days to obtain the final scheduling decision for each running day.

8. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps in the microgrid energy storage full life cycle optimized operation method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the microgrid energy storage full life cycle optimization operation method as described in any one of claims 1-6.

10. A computer program product, comprising software code, characterized in that, The program in the software code executes the steps in the microgrid energy storage full life cycle optimization operation method as described in any one of claims 1-6.

Images